developerjeremylive/NVILA-HD-Video-AutoGaze-etheroi
gradio asyncio patch
upgradio 6.11.0
upgradio 6.11.0
Improve ZeroGPU duration estimation using video metadata and cache-awareness for faster, more accurate allocation
Improve ZeroGPU duration estimation using precomputed video metadata and unify pipeline with cached video-aware scheduling
check video metadata before zerogpu estimation
Optimize video + preview caching to eliminate redundant recomputation and speed up chat turns.
back to workiness
feat: yield AutoGaze preview earlier in chat flow + improve loading UX
Update app.py
refactor: extract AutoGaze preview before chat + simplify streaming pipeline
Update README.md
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Fix chat worker streaming loop to handle streamer timeouts safely
Refactor chat turn with worker queue for continuous waiting animation and streamed replies
Show chat waiting dots immediately and keep them animated until first streamed token
Add waiting dots animation to chat assistant bubble before first streamed token
Fix chat behavior: display user message instantly and stream assistant response
Update app.py
Update app.py
add duplicate button
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Update README.md
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
Update app.py
fix(zerogpu): increase cold-start GPU allocation for first model load
fix(zerogpu): make duration estimator compatible with one-shot and chat handlers
add chatbot mode + perf improvements
clean up NVILA caching app with clearer status messages and reduced logging
Update app.py
remove experimental NVILA processor cache and keep AutoGaze preview disk cache
add disk cache for NVILA video preprocessing and reuse it across prompts
align AutoGaze preview with main tile gazing settings
switch AutoGaze preview cache from memory to disk
add cache for AutoGaze visualization video and metrics